Skip to main content

AI Platform Engineer path

For platform and infrastructure engineers who are asked to run models, not train them. The path takes the container and Kubernetes primitives you need (isolation, images and registries, the reconciler, the scheduler, Services and the Gateway API) and goes straight to the parts that are different for AI workloads: GPUs as a scarce and divisible resource, model servers as batch schedulers with a KV cache, gateways that count tokens, autoscaling on queue depth, prefix-aware routing, vector stores, models as supply-chain artifacts, and what a token really costs.

Lessons appear here as they are published; prerequisites are listed on each lesson page.

AI Platform Engineer · 23 lessons · 33.5 h
  1. Build The Lab: One Command That Proves Your Machine Is Ready · Phase 00 Setup & Lab · 60 min
  2. Build Process Isolation: A Container Is Just a Process · Phase 05 Containers · 90 min
  3. Learn Images Are Just Layers · Phase 05 Containers · 45 min
  4. Build Pull an Image With Plain HTTP: Registries Are Content-Addressed · Phase 05 Containers · 75 min
  5. Build Build an Image Without a Dockerfile: Layers, Config and Manifest by Hand · Phase 05 Containers · 90 min
  6. Build A Container's Network Is a veth Pair, a Bridge and Three NAT Rules · Phase 05 Containers · 90 min
  7. Build Compose Is a Dependency Graph Over docker run · Phase 05 Containers · 90 min
  8. Build A Controller Is a Loop That Compares: Reconcilers From Scratch · Phase 06 Orchestration · 90 min
  9. Build A Scheduler Is Filter, Score, Bind · Phase 06 Orchestration · 90 min
  10. Build A Node Agent Turns a Pod Spec Into Processes · Phase 06 Orchestration · 90 min
  11. Build A Service Is a Virtual IP and a Rewrite Rule · Phase 06 Orchestration · 90 min
  12. Build Routing Is Listeners, Routes and Precedence: the Gateway API · Phase 06 Orchestration · 90 min
  13. Build A GPU Is a Countable Resource Until It Isn't · Phase 13 AI Infrastructure · 90 min
  14. Build Serving a Model Is a Batch Scheduler With a KV Cache · Phase 13 AI Infrastructure · 120 min
  15. Build An AI Gateway Is a Reverse Proxy That Counts Tokens · Phase 13 AI Infrastructure · 90 min
  16. Build Scale Inference on Queue Depth, Not CPU · Phase 13 AI Infrastructure · 90 min
  17. Build Route by Prefix Cache, Not Round Robin · Phase 13 AI Infrastructure · 90 min
  18. Build A Vector Index Is a Recall Budget You Tune · Phase 13 AI Infrastructure · 90 min
  19. Build A Model Is an Artifact: Weights, Formats and Quantization · Phase 13 AI Infrastructure · 75 min
  20. Build The Cost of a Token Is a Capacity Plan · Phase 13 AI Infrastructure · 75 min
  21. Build An MCP Server for kubectl: Tools Are Functions With a Schema · Phase 14 AIOps & Agentic Ops · 90 min
  22. Build An Incident-Triage Agent: The Loop Decides What the Model May Touch · Phase 14 AIOps & Agentic Ops · 120 min
  23. Build Trace the Agent Like a Service: OpenTelemetry for LLM Calls · Phase 14 AIOps & Agentic Ops · 90 min

Pair this with the AI/ML Infrastructure career page and the AI/ML Ops learning path in the Reference layer.